Well, this week has been a very busy one for me, but I've learned a few
new things which should be of major interest to to anyone working with, or
contemplating working with, digital negatives.
Before I get into, I should mention that the service bureau I mentioned
in my last post -- Advance Reprographics in N. Andover, MA -- has NOT
been able to make good film for me yet. They are having some trouble
with their imagesetter. To their credit, they are still trying to
resolve their problems, however, at this time I would have to advise not
calling them.
Now, on with the show...
1. The digital negatives I've been running tests with appear to only have
a resolution of 150 lpi rather than the 300 lpi reported earlier. My
friend who output them for me swears they are 300 lpi, but two other
pre-press people and myself have confirmed (using a 100x louple with a
reticle) that they are really only 150 lpi. In the coming week, I will be
visiting with my friend so that he and I can double-check the calculations.
Anyway, the practical upshot is that I think prints made from those
negatives are actually quite good -- possibly good enough for many
people. Charlie Palmer and Pradip Malde, who both have samples of the
tests made from those negs can confirm this. Therefore, the quest for
outrageously high-resolution linescreens may not be necessary. In fact,
if I can double the number of dots per inch by going to 300 lpi, the
resulting print should look great! This also means that for some people,
it may now be possible to scan at lower resolutions, resulting in
smaller, more manageable files.
This brings me to the next topic...
2. As I suspected, the relationship between the frequency of the
linescreen used (in lpi) and the final resolution of your image file (in
ppi) necessary for zero data loss is governed by the Nyquist Sampling
Theorem (just like audio CDs). The Nyquist Theorem states that your
sampled rate (your scan resolution in ppi) needs to be 2 times the desired
output frequency (your linescreen in lpi). So don't believe it when all
these service bureaus tell you you only require 1.5 times the
resolution. That's what they use -- saves money, time and space -- and
it may be acceptable for them, but it does result in some data loss.
3. Here's the real stunner. Ready?
There is a formula which relates the frequency of the linescreen you are
using (in lpi), the resolution of the imagesetter (in dpi), AND the
number of shades of gray that your film will be capable of rendering.
The equation is:
__ __ 2
| |
| output resolution of the imagesetter in dpi |
| ___________________________________________ | + 1
| |
| frequency of the linescreen in lpi |
|__ __|
= the number of shades of gray that the negative can render
And, I have to tell you, the results are not pretty!
Here is a small taste...
| frequency of linescreen used
Res. of |
imagesetter | 200 lpi 300 lpi 400 lpi
____________|____________________________________________
|
2400 dpi | 145 65 36
|
3600 dpi | 325 145 81 Shades
| of
4000 dpi | 401 178 101 Gray
|
4800 dpi | 577 257 145
|
6400 dpi | 1025 456 257
I'm sure you will recall that most programs render grayscale images using
8 bits resulting in 256 shades of gray... So, my friends, the moral of
the story is that unless you have a final image file resolution of 800
ppi, and a service bureau with an imagesetter that can do 6400 dpi output,
then you are probably better off using a linescreen frequency lower than
400 lpi if you want to render all 256 shades of gray (sounds like a good
idea to me). Good thing that it now looks as though 300 lpi will work well.
Well, that's it for now. I now return you to your regularly scheduled
mailing list.
David Fokos